A SURVEY OF INTRUSION DETECTION SYSTEMS IN IOT: MACHINE LEARNING AND FEATURE SELECTION APPROACHES

Abstract

The Internet of Things (IoT) enables automation and seamless data flow by connecting billions of devices across multiple industries. However, this networked environment also poses serious security risks, including denial-of-service attacks, unauthorized access, and data leaks. Intrusion Detection Systems (IDS) are essential for protecting IoT networks by detecting and halting malicious activity. This paper offers a comprehensive analysis of IDS created for IoT environments, looking at advancements, challenges, and possible patterns from 2020 to 2024. Hybrid IDS models have been created by combining multiple strategies to improve detection accuracy, scalability, and efficacy. While lightweight IDS frameworks balance robust security performance with the resource limitations of IoT devices, feature selection techniques further optimize these systems by reducing computational overhead. With some frameworks nearly reaching peak performance, deep learning and machine learning techniques have shown remarkable accuracy, often surpassing 95%. However, critical metrics like F1-score, accuracy, precision, and recall need to be carefully considered in order to assess these systems. This work also compares its results with popular datasets, such as NSL-KDD, CICIDS2017, UNSW-NB15, and Bot-IoT, among others, to enable meaningful comparisons between various IDS models. Emerging research employs synthetic data creation techniques to address problems like class imbalance, even as domain-specific IDS solutions for cloud-based IoT, industrial, and medical applications gain popularity. Despite these developments, there are still significant challenges, including the need for real-time flexibility and computational overhead. Finally, future research will probably focus on combining state-of-the-art artificial intelligence techniques like supervised and unsupervised learning to create adaptive, lightweight, and privacy-preserving IDS for IoT environments.